Notebook

Meta Platforms

META Open Platforms

  • −19.0%unrealized, USD
  • 673.66average cost, as supplied
  • 545.83last · 2026-08-20

The record

Updated
2026-08-19.

Thesis

Written 2026-08-21.

About 3.6 billion people — roughly 44% of the planet — use something Meta owns every single day, and the machine that sells advertising against them is getting better, not just bigger. This is the position I feel most strongly about, and the market disagrees with me right now, which is most of why I hold it.

Ad volume and ad price, moving together

growth versus the same quarter a year earlier, per cent

Impressions Q1 2026 19% vs Q1 2025 Impressions Q2 2026 14% vs Q2 2025 Price per ad Q1 2026 12% vs Q1 2025 Price per ad Q2 2026 12% vs Q2 2025

Scroll the chart sideways →

What you're looking at: Ad impressions and average price per ad, each against the same quarter a year before. These two normally trade off against each other — showing more ads usually depresses the price of each. Both rising at once is the unusual result, and it is the single best evidence that the ad system is getting better rather than just bigger. The qualifier is in the first pair: impression growth slowed from 19% to 14% between the quarters, while pricing held. Meta Q1 and Q2 2026 reported results. Company-reported, not audited by me, not pre-registered. Source numbers: meta-ads.csv.
Show the numbers
Measure ValueNote
Impressions Q1 202619%vs Q1 2025
Impressions Q2 202614%vs Q2 2025
Price per ad Q1 202612%vs Q1 2025
Price per ad Q2 202612%vs Q2 2025

That chart is the whole argument in one image. Ad impressions rose 14% and the average price per ad rose 12% in the same quarter. Those two normally trade against each other — showing more ads usually pushes down what each is worth. Both rising together means the targeting is genuinely improving. Meta does not break out an "AI revenue" line, so this is where the AI shows up in the accounts, and it is the reason the stock has been sold harder than peers who do quantify it.

Revenue growth, nine quarters in a row

total revenue versus the same quarter a year earlier, per cent

0 10 20 30 19% Q3 2024 21% Q4 2024 16% Q1 2025 low 22% Q2 2025 26.3% Q3 2025 23.8% Q4 2025 33.1% Q1 2026 peak 28% Q2 2026 22% Q3 2026 guidance

Scroll the chart sideways →

What you're looking at: Nine consecutive quarters of revenue growth, each measured against the same quarter a year before. This is the clearest way I can show what I mean by acceleration: growth bottomed at 16% in the first quarter of 2025, and every quarter since has come in above it, peaking at 33.1%. The business is now growing faster than it was two years ago while being roughly a third larger. Two honest qualifiers. It is not a smooth line - Q4 2025 dipped, and the latest quarter has come off the peak. And the final bar is the company's own guidance rather than a result, drawn open for that reason, implying about 22% at the midpoint. I have started the chart in 2024 on purpose: beginning it at the 16% low would have made the run look cleaner than it is. Meta company-reported total revenue growth. Q3 2024 $40.59bn (+19%), Q4 2024 $48.39bn (+21%), Q1 2025 $42.31bn (+16%), Q2 2025 $47.52bn (+22%), Q3 2025 $51.24bn, Q4 2025 $59.89bn, Q1 2026, Q2 2026 $60.8bn. Q3 2026 is guidance of $61-64bn shown at the midpoint against reported Q3 2025 revenue. Rates are as the company reported them, to the precision available for each quarter. Company-reported, not audited by me, not pre-registered. Source numbers: meta-revenue-growth.csv.
Show the numbers
Period ValueNote
Q3 202419%
Q4 202421%
Q1 202516%low
Q2 202522%
Q3 202526.3%
Q4 202523.8%
Q1 202633.1%peak
Q2 202628%
Q3 202622%guidance

Growth has roughly doubled off its 2025 low, and stayed there. It bottomed at 16% in Q1 2025 and every quarter since has come in above that, peaking at 33.1% and standing at 28% on $60.8bn of quarterly revenue. Meta is growing faster today than it was two years ago while carrying a base roughly a third larger. That is not a mature business being harvested.

It is about to become the largest advertising business in the world. Meta is forecast to take $243.5bn of net worldwide ad revenue in 2026 against Google's $239.5bn — 26.8% of global ad spend against 26.4% — passing Google for the first time. The reason is the growth gap: roughly 24% against about 12%. The Advantage+ automated suite is named as the single biggest driver, which is the same AI story as the chart above, arriving as market share. It is a full-year forecast rather than a reported result, but the direction is not in doubt.

On the capital expenditure, I take the other side of the sell-off, and I think the sellers have it backwards. Meta has been short of compute and renting it. Spending to own that capacity is a company refusing to depend on anyone else — and it has learned that lesson expensively before. On the February 2022 earnings call, CFO Dave Wehner put Apple's tracking changes at a headwind "on the order of $10 billion" of revenue in a single year. Zuckerberg has estimated the company "might be twice as profitable if we own the platform". Both are management's own estimates rather than audited disclosures, and the $10bn figure is now four years old — but they are the numbers the people running the company used, and they explain exactly why it would rather own compute than rent it.

This is the one page where the comparison with the other two earns its place, so here are all three.

What they said they would spend, and what they say now

capital expenditure planned for 2026, billions of dollars

Amazon — said earlier $200bn Amazon — says now $220bn $20bn more Alphabet — said earlier $185bn Alphabet — says now $205bn $20bn more Meta — said earlier $135bn Meta — says now $137bn $2.5bn more

Scroll the chart sideways →

What you're looking at: Each company appears twice. The first bar is what it told investors earlier in the year. The second is what it tells them now. Nothing else changed in between except the plan. Amazon added $20bn, Alphabet added about $20bn, Meta about $2.5bn. Read the pairs, not the heights: the point is the gap inside each pair, which is how much more than expected each is now spending. If the build-out were settling down, the second bar in each pair would be the shorter one. Company guidance. Amazon $200bn raised to $220bn, attributed partly to higher memory prices. Alphabet $180-190bn raised to as much as $205bn, shown at $185bn and $205bn. Meta $125-145bn narrowed to $130-145bn, shown at the midpoints of $135bn and $137.5bn. Company-reported guidance, not audited by me, not pre-registered. Source numbers: hyperscaler-capex-2026.csv.
Show the numbers
Measure ValueNote
Amazon — said earlier$200bn
Amazon — says now$220bn$20bn more
Alphabet — said earlier$185bn
Alphabet — says now$205bn$20bn more
Meta — said earlier$135bn
Meta — says now$137bn$2.5bn more

Meta is not the biggest spender of the three, and it raised its plan by the least - about $2.5bn, against roughly $20bn each for Amazon and Alphabet. It is the one being punished hardest for it. The gap between what the company is actually doing and how the stock has been treated is why this is a position rather than a watchlist entry.

A four-billion-person network with the best ad targeting in the world, printing operating cash and buying its own future, is not a business I want to trade around.

What I am watching

I am bullish, and these are the things I would need to see to change that.

The litigation, which is the threat I take most seriously — though not for the reason the headline suggests. The $1.4tn figure in circulation is a theoretical ceiling on civil penalties, and the states have said they are not seeking a specific amount; reading it as the risk gets this wrong. What is actually in front of the company: a jury found Meta liable in March 2026, one case has produced $942m in awards, and the federal trial with California, Colorado, Kentucky and New Jersey opened in August 2026, with twenty-six more states behind them.

A fine is a number a company this size absorbs. The exposure that would matter is a product remedy — the claims are about design, meaning infinite scroll, algorithmic recommendation and the like button. A court-ordered change there would not be scoped to teenagers; it would reach the engagement machinery the whole ad system is priced on. That is the branch I am watching, and I judge it unlikely rather than impossible.

Growth against the guide. Measured from the Q1 2025 low this is an acceleration; measured from the Q1 2026 peak it is not, and the company's own Q3 guidance implies about 22% at the midpoint. Impressions tell the same story, 19% in Q1 against 14% in Q2, with pricing flat at 12%. The AI improvement is visible in both charts, and so is the fact that it has a limit.

Cash conversion. Capital expenditure of $31.1bn consumed roughly 98% of operating cash flow, leaving $784m of free cash flow against $8.55bn a year earlier, with capex including finance leases running at 41% of revenue. I think this is the right thing to be doing. I still want to see it stop.

One argument I am deliberately not making. It is tempting to note that buying Meta at multiples like these has worked out before. I am leaving that off the page: it reasons from the outcome backwards, it would have said the same thing at points where it was wrong, and a record whose whole value is that it was written in advance cannot lean on hindsight.

What would make me wrong

  • Price per ad flattening while impression growth keeps decelerating — the two halves failing at once.
  • A product remedy that reaches the core recommendation system, rather than a fine.
  • Capex rising again into 2027 with no revenue acceleration to match.
  • Free cash flow near zero for several more quarters.

Latest movements

  • 2026-07-15 — Opened in two lots, at $675.25 and $684.00.
  • 2026-07-17 — Added a top-up at $648.19. Position size +20%.
  • 2026-08-06 — Trimmed 33% at $589.66, cut at a loss. Realized −12.5%.